[lammps-users] airebo forces

LAMMPS LAMMPS Mailing List Mirror
1

hi all,

I'm having trouble using the airebo potential.

so if i switch off the LJ term everything is fine, i get decent forces, if i switch it on, forces go crazy. i attached my script and datafile.

Thanks very much in advance.
Albert

>>
units metal

atom_style atomic
read_data data.lammps

pair_style airebo 3.0 0 1
pair_coeff * * CH.airebo C

dump 1 all custom 1 force.lammps id x y z fx fy fz
dump_modify 1 format "%d %20.15g %20.15g %20.15g %20.15g %20.15g %20.15g"

run 0
<<

data.lammps
>>

   48 atoms
   1 atom types

   0.000000 5.712550 xlo xhi
   0.000000 5.034600 ylo yhi
   0.000000 8.720183 zlo zhi
   0.000000 0.000000 0.000000 xy xz yz
   Masses

   1 1.0

   Atoms

   1 1 3.655415 2.509590 6.707051
   2 1 3.716075 3.741860 4.517012
   3 1 3.579865 3.686960 7.377382
   4 1 2.032075 3.904030 7.746562
   5 1 4.035645 4.833370 2.347942
   6 1 3.532895 1.175300 4.613641
   7 1 4.061155 4.964600 6.765902
   8 1 3.824305 4.887920 5.304862
   9 1 1.628405 0.117090 5.372552
   10 1 3.575635 1.116340 7.495992
   11 1 2.006055 1.425460 7.683511
   12 1 3.748975 1.099420 0.153792
   13 1 3.986905 3.460840 0.288412
   14 1 4.012755 2.202670 2.271102
   15 1 3.291605 2.253580 0.921902
   16 1 1.985135 2.609650 0.926052
   17 1 3.537125 0.957290 3.170852
   18 1 1.827615 1.540290 3.129371
   19 1 3.889205 3.572190 3.063802
   20 1 2.273855 3.739110 3.195842
   21 1 3.618265 2.393570 5.154371
   22 1 1.700635 2.799600 5.360191
   23 1 1.677685 2.723160 3.861232
   24 1 5.708565 2.867490 3.617472
   25 1 1.971405 1.603000 6.025212
   26 1 5.688235 1.666830 5.861422
   27 1 1.877485 3.965540 6.008612
   28 1 5.586855 3.959910 5.993082
   29 1 0.002375 2.844890 6.731922
   30 1 1.773975 3.894750 1.897732
   31 1 0.255345 3.934260 1.774212
   32 1 5.687425 4.173030 4.611382
   33 1 3.353515 4.777940 1.068352
   34 1 1.732385 0.261730 1.029062
   35 1 1.476715 0.231590 8.337012
   36 1 5.698175 0.128300 8.253542
   37 1 1.515185 1.528190 1.782002
   38 1 5.703035 1.253440 1.608272
   39 1 1.895855 0.010550 3.897442
   40 1 5.638115 0.399790 3.848842
   41 1 5.427295 0.162060 2.347922
   42 1 5.621725 1.615640 4.369421
   43 1 5.597095 0.405720 6.793342
   44 1 1.478695 2.672120 8.330602
   45 1 0.009585 2.679960 8.358911
   46 1 5.159525 1.405480 0.101342
   47 1 5.353095 3.958250 0.341922
   48 1 5.437375 2.685880 2.210432
<<

2

Could be any number of things: bad geometry,
bad timestep, etc. Can you relax your system,
are the dynamics stable? Can you reproduce
some AIREBO calculation from a paper?

Steve

3

Hi Steve,

The geometry seems reasonable to me, but that's not the point. The REBO
potential should be a reasonable description for carbon atoms on its
own, and indeed, if I switch off the LJ term, forces are good, similar
to those I get with another carbon potential (ie Tersoff) or even DFT.
That shows that the geometry is sensible (it's also a self-consitency
check for LAMMPS). Switching the LJ term on should make too much
difference, as it adds only a bit of dispersion interaction to the
system, maybe a small fraction of the REBO forces. Here, I got huge
forces, 10-20 times larger than the REBO forces. I don't intend to do
relaxation or dynamics, (I don't see any sensible results coming out
until the forces show this behaviour) I just want the forces in this
single geometry sorted first.

I agree that I must be doing something wrong but I can't figure out
what. I haven't made any changes in the potential file neither in the
code and tried to follow the instruction in the manual closely. If yoy
can offer any help, it would be much appreciated.

Cheers,
Albert

Steve Plimpton wrote:

4

hi, Albert

Where is your LJ option in the input script? what do you mean turn off the lJ term? you are using the hybrid potential too? Maybe we are asking the same question but from different angle.

Best,

AC

5

Hi Ajing,

in the input script the line

pair_style airebo 3.0 0 1

controls the terms included in the AIREBO potential. the three numbers
after the keyword 'airebo' are the following:
3.0 is the cutoff of the LJ term in sigma units
0 is that LJ term is switched off (changing to 1 switches it on)
1 is the switch for torsion, that hasn't caused trouble to me.

So if you run lammps with my input script, you get a bunch of forces
that look okay (as I said, compared to other potentials including DFT),
but if you change that 0 to 1, forces go crazy, you can see that by
simply inspecting the force.lammps file.

Cheers,
Albert

AJING CAO wrote:

6

Just b/c the LJ forces are larger than the REBO forces
doesn't mean much to me. If your geometry has
surfaces or is inhomogenous in some way, how do
you know the forces shouldn't be larger? I would try
to compare to some published geometry or energies.
If you think there is a problem with AIREBO in LAMMPS,
then I'd email Ase Henry (ase at mit.edu) and ask him
your question.

Steve